Deliberative Wildfire Networks
Lisa F. Oswald
November 16, 2021 | 13:27:46 | CET
Comments were collected using the PRAW, a Python wrapper for a Reddit API. The comment text was collected together with the reply-to structure (identified through comment ids and parent-ids). Using this procedure, a network structure of nodes (comments) and edges (reply-to-structure) was constructed. Discussion (sub)threads were identified using the network metric of weakly connected components. Within one API call, PRAW collects either comments or submissions, not both belonging together, therefore these discussion threads are considered as sub-treads, starting with root comments, comments directly commenting a reddit submission or post.
Another API call using PRAW collected the submissions that triggered the root-comments. They were traced back using the parent-id of the root comments. In other words, the mothers for all the small comment threads had to be identified in order to merge sibling-threads, belonging to one original trigger-mother-submission, together.
These trigger submissions were manually coded with qualitative content analysis. A code of 1 indicates content that expresses skepticism about the relationship between climate change and wildfires, while a code of 0 indicates the opposite: concern about this very relation. A code of 2 indicates that the content does not directly relate to wildfires but that e.g. the word “fire” was used in a metaphorical way (e.g. “Someone is on fire.”). The following tables display the summary statistics for the discussion thread characteristics of the whole threads, the sub-treads and the comments.
In a manual step, the thread / sub-thread matching was checked.
## Data preparation
library(tidyverse)
library(ggplot2)
library(readr)
library(readxl)
library(xtable)
library(knitr)
library(kableExtra)
library(writexl)
data_cc_networks <- read_excel("../../data/final/data_cc_networks_coded.xlsx")
## get depth metric for overall (big) thread
data_cc_networks <- data_cc_networks %>%
group_by(id_sub)%>%
mutate(thread_depth = max(depths))
## get with metric for overall (big) thread
data_cc_networks <- data_cc_networks %>%
group_by(id_sub)%>%
mutate(thread_width = max(widths))
## compress data set to trigger submissions only
threads <- data_cc_networks%>%
filter(type_sub != "is.na")
## load comment level data
reddit_df_indegree <- read_csv("../../data/raw/reddit_df_indegree_thread_id.csv")
## merge comment data with trigger submissions
fire_data <- merge(x = threads, y = reddit_df_indegree, by.x = "id_rc", by.y = "id" , all = TRUE)
fire_data <- fire_data%>%
group_by(subreddit, wave.y)%>%
arrange(thread_id)
# help with sorting subthreads to submissions
fire_data <- fire_data%>%
group_by(subreddit_y, wave.y, thread_id)%>%
mutate(sub_thread_id = group_indices())
#write_xlsx(fire_data,"data/temp/fire_data.xlsx")
# manual match contolling & thread sorting
#### creating final dataset #####
data <- read_excel("../../data/temp/fire_data_clean.xlsx")
# opposing variables
data <- data%>%
mutate(opposing = ifelse((type_sub == 1 & subreddit == "climatechange"), 1,
ifelse((type_sub == 0 & subreddit == "climateskeptics"), 1, 0)))
#write_xlsx(data,"data/temp/data_fire_cc.xlsx")Sample
data <- read_excel("../../data/temp/data_fire_cc.xlsx")
#data <- data.frame(na.omit(data))
data <- data%>%
mutate(type_sub = recode(type_sub, "3" = 1))
# 1543 comments
length(unique(data$id_c))
# written by 572 unique authors
length(unique(data$author_c))
# 524 sub-threads
length(unique(data$thread_id))
# 115 threads
length(unique(data$id_sub))
# 62 unique authors writing the submissions
length(unique(data$author_sub))
#### opposing content #####
opposing_data <- data%>%
filter(opposing == 1)
# 168 / 1543 opposing comments
length(unique(opposing_data$id_c))
# written by 93 authors
length(unique(opposing_data$author_c))
# 86 / 524 sub-threads
length(unique(opposing_data$thread_id))
# 21 / 115 threads
length(unique(opposing_data$id_sub))
# written by 16 authors
length(unique(opposing_data$author_sub))
# unrelated comments
table(data$type_sub)
thread_data <- data %>%
group_by(id_sub) %>% slice(1)
table(thread_data$type_sub)NAs kept
The final dataset consists of 1543 comments, written by written by 572 different authors. These comments belong to 524 sub-threads and 115 threads. The trigger submission for the 115 threads were written by 62 unique authors. 168 out of the 1543 comments followed 21 submissions with opposing content relative to the subreddit’s mainstream content. These opposing submission were written by 16 different authors. In total, 7 threads did not contain discussions on the 2020 wildfires but e.g. used the word ‘fire’ in a metaphorical way. They were excluded from the analyses.
NAs removed
The final dataset consists of 1404 comments, written by written by 543 different authors. These comments belong to 472 sub-threads and 108 threads. The trigger submission for the 108 threads were written by 61 unique authors. 150 out of the 1404 comments followed 20 submissions with opposing content relative to the subreddit’s mainstream content. These opposing submission were written by 15 different authors. In total, 6 threads including 60 comments did not contain discussions on the 2020 wildfires but e.g. used the word ‘fire’ in a metaphorical way. They were excluded from the dataset.
# split waves
data_cal <- data%>%
filter(wave == "california")
data_aus <- data%>%
filter(wave == "australia")Summary Statistics for different Discussion Threads
Ausralia Data
options(digits = 3)
data_aus %>%
group_by(subreddit, opposing)%>%
summarize(max_depth = max(max_thread_depth), mean_depth = mean(max_thread_depth),
max_widht = max(mean_thread_width), mean_width = mean(mean_thread_width),
max_ups = max(ups_sub), mean_ups = mean(ups_sub),
max_comms = max(n_comments_sub), mean_coms = mean(n_comments_sub)) %>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | opposing | max_depth | mean_depth | max_widht | mean_width | max_ups | mean_ups | max_comms | mean_coms |
|---|---|---|---|---|---|---|---|---|---|
| climatechange | 0 | 28 | 10.43 | 0.980 | 0.911 | 143 | 90.1 | 74 | 36.8 |
| climatechange | 1 | 6 | 4.22 | 0.833 | 0.715 | 62 | 18.4 | 19 | 14.9 |
| climatechange | NA | NA | NA | NA | NA | NA | NA | NA | NA |
| climateskeptics | 0 | 13 | 6.55 | 0.962 | 0.761 | 118 | 65.1 | 30 | 13.9 |
| climateskeptics | 1 | 23 | 23.00 | 0.964 | 0.964 | 0 | 0.0 | 32 | 32.0 |
California Data
# create summary tables for deliberative structure estimates
data_cal %>%
group_by(subreddit, opposing)%>%
summarize(max_depth = max(max_thread_depth), mean_depth = mean(max_thread_depth),
max_widht = max(mean_thread_width), mean_width = mean(mean_thread_width),
max_ups = max(ups_sub), mean_ups = mean(ups_sub),
max_comms = max(n_comments_sub), mean_coms = mean(n_comments_sub)) %>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | opposing | max_depth | mean_depth | max_widht | mean_width | max_ups | mean_ups | max_comms | mean_coms |
|---|---|---|---|---|---|---|---|---|---|
| climatechange | 0 | 8 | 5.21 | 0.909 | 0.812 | 136 | 75.18 | 22 | 15.4 |
| climatechange | 1 | 8 | 5.00 | 0.909 | 0.701 | 13 | 3.32 | 16 | 12.5 |
| climateskeptics | 0 | 25 | 8.56 | 0.971 | 0.837 | 254 | 127.17 | 97 | 44.6 |
| climateskeptics | 1 | 3 | 2.46 | 0.750 | 0.591 | 11 | 3.42 | 19 | 10.9 |
| climateskeptics | NA | NA | NA | NA | NA | NA | NA | NA | NA |
Group sizes
Number of Comments
data%>%
group_by(subreddit, wave)%>%
summarise(n())%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | n() |
|---|---|---|
| climatechange | australia | 598 |
| climatechange | california | 170 |
| climateskeptics | australia | 126 |
| climateskeptics | california | 648 |
| NA | NA | 194 |
Number of Threads
thread_data%>%
group_by(subreddit, wave)%>%
summarise(n())%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | n() |
|---|---|---|
| climatechange | australia | 30 |
| climatechange | california | 18 |
| climateskeptics | australia | 16 |
| climateskeptics | california | 51 |
Different Width estimates
- Mean in-degrees of comments per thread
- Number of different users engaged in thread
- Maximum width of thread at any given depth level
# Mean in-degree
ggplot()+
geom_histogram(aes(data$mean_thread_width))+
xlab("Mean in-degrees of comments per thread")
# Unique authors
data <- data%>%
dplyr::group_by(id_sub)%>%
dplyr::mutate(unique_authors = n_distinct(author_c))
ggplot()+
geom_histogram(aes(data$unique_authors))+
xlab("Number of unique authors engaged in thread")
# Maximum width (number of comments) on a given depth level
data <- data%>%
group_by(id_sub, depth_c)%>%
mutate(width_values = sum(in_degree_com))%>%
group_by(id_sub)%>%
mutate(gonzalez_width = max(width_values))
ggplot()+
geom_histogram(aes(data$gonzalez_width))+
xlab("Maximum width of thread at any given depth level")
Summary statistics - comment level
# drop NAs
data <- data.frame(na.omit(data)) ### impute NAs to keep full sample??One node represents one comment, in-degrees represent indegrees of comments (how many replies did one comment get), ups refer to comments (how many ups did one comment get)
# plot in_degree distributions
ggplot(data, aes(n_comments_sub, mean_thread_width))+
geom_point(aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 10, height = 0.08),
alpha = 0.5)+
geom_point(data = subset(data, opposing == 1), aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 10, height = 0.08),
alpha = 0.5)+
theme_minimal()+
facet_grid(vars(wave), vars(subreddit))+
xlab("Number of Comments")+
ylab("Width (in-degree)")+
scale_colour_manual(name="Type",
labels=c("congruent", "opposing"),
values=c("#2E8B57", "indianred")) +
scale_size_continuous(name="Up-votes")
# plot in_degree distributions
ggplot(data, aes(max_thread_depth, mean_thread_width))+
geom_point(aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 6, height = 0.08),
alpha = 0.5)+
geom_point(data = subset(data, opposing == 1), aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 6, height = 0.08),
alpha = 0.5)+
theme_minimal()+
facet_grid(vars(wave), vars(subreddit))+
xlab("Depth")+
ylab("Width (in-degree)")+
scale_colour_manual(name="Type",
labels=c("congruent", "opposing"),
values=c("#2E8B57", "indianred")) +
scale_size_continuous(name="Up-votes")
ggplot(data, aes(max_thread_depth, mean_thread_width))+
geom_point(aes(colour = as.factor(opposing), size = n_comments_sub),
#position = position_jitter(width = 3, height = 0.08),
alpha = 0.5)+
geom_point(data = subset(data, opposing == 1),
aes(colour = as.factor(opposing), size = ups_sub),
äposition = position_jitter(width = 3, height = 0.08),
alpha = 0.5)+
#theme_grey()+
facet_grid(vars(wave), vars(subreddit))+
xlab("Depth")+
ylab("Width (in-degree)")+
scale_colour_manual(name="Type",
labels=c("congruent", "opposing"),
values=c("#2E8B57", "indianred")) +
scale_size_continuous(name="N comments")
Hpyotheses Tests
H1: Threads in reaction to opposing views have lower scores.
par(mar = c(4, 4, .1, .1))
# comment level data
data$opposing <- as.factor(data$opposing)
my.mean<-tapply(data[,"ups_sub"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"ups_sub"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 102.7 | 67.5 | 1.91 |
| 1 | opposing | 10.6 | 20.5 | 1.67 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("indianred")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=15),
axis.title.y = element_text( colour="black", size=15))+
labs(x="",y="mean up-votes of submissions",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
my.mean<-tapply(thread_data[,"ups_sub"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"ups_sub"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 62.70 | 57.9 | 5.97 |
| 1 | opposing | 6.45 | 13.7 | 3.05 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("indianred")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=15),
axis.title.y = element_text( colour="black", size=15))+
labs(x="",y="mean up-votes of submissions", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(ups_sub)/sqrt(length(ups_sub)),
mean = mean(ups_sub))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 11.04 | 64.0 |
| climatechange | australia | 1 | 10.04 | 12.2 |
| climatechange | australia | NA | NA | NA |
| climatechange | california | 0 | 12.89 | 55.1 |
| climatechange | california | 1 | 1.83 | 3.2 |
| climateskeptics | australia | 0 | 9.59 | 36.5 |
| climateskeptics | australia | 1 | NA | 0.0 |
| climateskeptics | california | 0 | 10.22 | 73.5 |
| climateskeptics | california | 1 | 1.64 | 5.0 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("indianred2", "indianred4"))+
labs(x="opposing submission content",y="mean up-votes of submissions", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$ups_sub ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$ups_sub by thread_data$opposing
t = 8, df = 111, p-value = 2e-13
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
43.0 69.5
sample estimates:
mean in group 0 mean in group 1
62.70 6.45 ANOVA for opposing vs. congruent x wave
# overall t-test
summary(aov(ups_sub ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 52183 52183 18.57 3.6e-05 ***
wave 1 4148 4148 1.48 0.23
opposing:wave 1 1865 1865 0.66 0.42
Residuals 110 309172 2811
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
1 observation deleted due to missingnessANOVA for opposing vs. congruent x ideology (subreddit)
# overall t-test
summary(aov(ups_sub ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 52183 52183 18.23 4.2e-05 ***
subreddit 1 100 100 0.04 0.85
opposing:subreddit 1 188 188 0.07 0.80
Residuals 110 314896 2863
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
1 observation deleted due to missingness3-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
# overall t-test
summary(aov(ups_sub ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 52183 52183 18.51 3.8e-05 ***
subreddit 1 100 100 0.04 0.851
wave 1 4316 4316 1.53 0.219
opposing:subreddit 1 231 231 0.08 0.775
opposing:wave 1 1596 1596 0.57 0.453
subreddit:wave 1 9513 9513 3.37 0.069 .
opposing:subreddit:wave 1 596 596 0.21 0.647
Residuals 106 298832 2819
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
1 observation deleted due to missingnessH2: Threads in reaction to opposing views have higher number of comments.
par(mar = c(4, 4, .1, .1))
# comment level data
data$opposing <- as.factor(data$opposing)
my.mean<-tapply(data[,"n_comments_sub"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"n_comments_sub"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 35.4 | 25.53 | 0.721 |
| 1 | opposing | 15.2 | 8.68 | 0.709 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("goldenrod")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="mean number of comments to submissions",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
my.mean<-tapply(thread_data[,"n_comments_sub"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"n_comments_sub"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 19.47 | 22.46 | 2.32 |
| 1 | opposing | 8.85 | 8.55 | 1.91 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("goldenrod")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="mean number of comments to submissions", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(n_comments_sub)/sqrt(length(n_comments_sub)),
mean = mean(n_comments_sub))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 4.05 | 24.22 |
| climatechange | australia | 1 | 2.79 | 11.67 |
| climatechange | australia | NA | NA | NA |
| climatechange | california | 0 | 2.09 | 10.46 |
| climatechange | california | 1 | 3.20 | 6.80 |
| climateskeptics | australia | 0 | 1.96 | 7.00 |
| climateskeptics | australia | 1 | NA | 32.00 |
| climateskeptics | california | 0 | 4.23 | 24.00 |
| climateskeptics | california | 1 | 2.05 | 5.12 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("goldenrod", "gold"))+
labs(x="opposing submission content",y="mean number of comments to submissions", title="Thread level data")+
facet_grid(vars(wave))
H3: Threads in reaction to opposing views are deeper.
par(mar = c(4, 4, .1, .1))
# comment level data
data$opposing <- as.factor(data$opposing)
my.mean<-tapply(data[,"max_thread_depth"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"max_thread_depth"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 8.23 | 4.25 | 0.120 |
| 1 | opposing | 6.45 | 6.78 | 0.554 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("skyblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="maximum depth of threads",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
my.mean<-tapply(thread_data[,"max_thread_depth"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"max_thread_depth"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 5.61 | 5.11 | 0.527 |
| 1 | opposing | 3.70 | 4.92 | 1.100 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("skyblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="maximum depth of threads", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(max_thread_depth)/sqrt(length(max_thread_depth)),
mean = mean(max_thread_depth))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 1.273 | 7.26 |
| climatechange | australia | 1 | 0.803 | 3.33 |
| climatechange | australia | NA | NA | NA |
| climatechange | california | 0 | 0.679 | 4.00 |
| climatechange | california | 1 | 1.378 | 3.00 |
| climateskeptics | australia | 0 | 0.875 | 3.73 |
| climateskeptics | australia | 1 | NA | 23.00 |
| climateskeptics | california | 0 | 0.824 | 5.86 |
| climateskeptics | california | 1 | 0.267 | 2.00 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("skyblue", "slateblue"))+
labs(x="opposing submission content",y="maximum depth of threads", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$max_thread_depth ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$max_thread_depth by thread_data$opposing
t = 2, df = 28, p-value = 0.1
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.592 4.405
sample estimates:
mean in group 0 mean in group 1
5.61 3.70 ANOVA for opposing vs. congruent x wave
summary(aov(max_thread_depth ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 60 59.9 2.33 0.13
wave 1 27 26.9 1.05 0.31
opposing:wave 1 42 41.7 1.63 0.21
Residuals 110 2824 25.7
1 observation deleted due to missingnessANOVA for opposing vs. congruent x ideology (subreddit)
summary(aov(max_thread_depth ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 60 59.9 2.29 0.13
subreddit 1 5 4.8 0.19 0.67
opposing:subreddit 1 15 15.0 0.57 0.45
Residuals 110 2873 26.1
1 observation deleted due to missingness3-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
summary(aov(max_thread_depth ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 60 60 2.71 0.103
subreddit 1 5 5 0.22 0.641
wave 1 22 22 1.00 0.319
opposing:subreddit 1 16 16 0.72 0.399
opposing:wave 1 84 84 3.78 0.054 .
subreddit:wave 1 26 26 1.15 0.285
opposing:subreddit:wave 1 399 399 18.05 4.6e-05 ***
Residuals 106 2342 22
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
1 observation deleted due to missingnessH4: Threads in reaction to congruent views are wider.
par(mar = c(4, 4, .1, .1))
# comment level data
data$opposing <- as.factor(data$opposing)
my.mean<-tapply(data[,"mean_thread_width"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"mean_thread_width"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 0.856 | 0.182 | 0.005 |
| 1 | opposing | 0.707 | 0.282 | 0.023 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="mean widdth of threads",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
my.mean<-tapply(thread_data[,"mean_thread_width"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"mean_thread_width"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 0.644 | 0.354 | 0.037 |
| 1 | opposing | 0.465 | 0.374 | 0.084 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="mean widdth of threads", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
thread_data <- data.frame(thread_data)
thread_data$opposing <- as.factor(thread_data$opposing)
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(mean_thread_width)/sqrt(length(mean_thread_width)),
mean = mean(mean_thread_width))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 0.067 | 0.757 |
| climatechange | australia | 1 | 0.170 | 0.536 |
| climatechange | australia | NA | NA | NA |
| climatechange | california | 0 | 0.087 | 0.661 |
| climatechange | california | 1 | 0.210 | 0.342 |
| climateskeptics | australia | 0 | 0.101 | 0.495 |
| climateskeptics | australia | 1 | NA | 0.964 |
| climateskeptics | california | 0 | 0.055 | 0.631 |
| climateskeptics | california | 1 | 0.099 | 0.427 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("steelblue2", "steelblue4"))+
labs(x="opposing submission content",y="mean widdth of threads", title="Thread level data")+
facet_grid(vars(wave))
Deliberative Structure
Creating measure for further analyses / comparison with deliberative quality: deliberative structure. mean_width x max_depth
data <- data%>%
mutate(del_structure = max_thread_depth*mean_thread_width)par(mar = c(4, 4, .1, .1))
# comment level data
my.mean<-tapply(data[,"del_structure"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"del_structure"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 7.55 | 4.34 | 0.123 |
| 1 | opposing | 5.59 | 6.80 | 0.556 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("seagreen")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="deliberative structure of threads",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data %>%
group_by(id_sub) %>% slice(1)
thread_data <- data.frame(thread_data)
my.mean<-tapply(thread_data[,"del_structure"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"del_structure"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 4.67 | 4.73 | 0.501 |
| 1 | opposing | 2.76 | 5.11 | 1.172 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("seagreen")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="deliberative structure of threads", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(del_structure)/sqrt(length(del_structure)),
mean = mean(del_structure))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 0.975 | 5.76 |
| climatechange | australia | 1 | 0.994 | 2.31 |
| climatechange | california | 0 | 0.686 | 3.16 |
| climatechange | california | 1 | 1.435 | 2.10 |
| climateskeptics | australia | 0 | 1.017 | 3.20 |
| climateskeptics | australia | 1 | NA | 22.18 |
| climateskeptics | california | 0 | 0.843 | 5.05 |
| climateskeptics | california | 1 | 0.285 | 1.03 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("seagreen", "seagreen3"))+
labs(x="opposing submission content",y="deliberative structure of threads", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$del_structure ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$del_structure by thread_data$opposing
t = 1, df = 25, p-value = 0.1
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.72 4.53
sample estimates:
mean in group 0 mean in group 1
4.66 2.76 ANOVA for opposing vs. congruent x wave
summary(aov(del_structure ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 57 56.8 2.50 0.12
wave 1 16 16.1 0.71 0.40
opposing:wave 1 56 56.0 2.46 0.12
Residuals 104 2364 22.7 ANOVA for opposing vs. congruent x ideology (subreddit)
summary(aov(del_structure ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 57 56.8 2.43 0.12
subreddit 1 0 0.4 0.02 0.90
opposing:subreddit 1 6 6.5 0.28 0.60
Residuals 104 2430 23.4 3-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
summary(aov(del_structure ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 57 57 2.92 0.090 .
subreddit 1 0 0 0.02 0.887
wave 1 21 21 1.09 0.299
opposing:subreddit 1 7 7 0.34 0.560
opposing:wave 1 89 89 4.58 0.035 *
subreddit:wave 1 7 7 0.35 0.554
opposing:subreddit:wave 1 368 368 18.92 3.3e-05 ***
Residuals 100 1944 19
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Further Ideas
Alternative Measure for width (representation) –> number of unique users involved
ggplot(data, aes(max_thread_depth, unique_authors))+
geom_point(aes(colour = as.factor(opposing), size = n_comments_sub),
position = position_jitter(width = 2, height = 5),
alpha = 0.5)+
geom_point(data = subset(data, opposing == 1),
aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 2, height = 5),
alpha = 0.5)+
#theme_grey()+
facet_grid(vars(wave), vars(subreddit))+
xlab("Depth")+
ylab("Width (engaged users)")+
scale_colour_manual(name="Type",
labels=c("congruent", "opposing"),
values=c("#2E8B57", "indianred")) +
scale_size_continuous(name="N comments")
par(mar = c(4, 4, .1, .1))
# comment level data
my.mean<-tapply(data[,"unique_authors"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"unique_authors"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 17.47 | 11.59 | 0.327 |
| 1 | opposing | 9.03 | 5.08 | 0.415 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="Number of different authors involved in threads",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data %>%
group_by(id_sub) %>% slice(1)
thread_data <- data.frame(thread_data)
my.mean<-tapply(thread_data[,"unique_authors"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"unique_authors"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 8.38 | 8.30 | 0.88 |
| 1 | opposing | 5.47 | 4.95 | 1.14 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="Authors involved in threads", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(unique_authors)/sqrt(length(unique_authors)),
mean = mean(unique_authors))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 1.74 | 11.45 |
| climatechange | australia | 1 | 2.27 | 7.60 |
| climatechange | california | 0 | 1.49 | 7.46 |
| climatechange | california | 1 | 2.46 | 4.60 |
| climateskeptics | australia | 0 | 1.02 | 4.62 |
| climateskeptics | australia | 1 | NA | 6.00 |
| climateskeptics | california | 0 | 1.49 | 8.37 |
| climateskeptics | california | 1 | 1.81 | 4.62 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("steelblue4", "steelblue2"))+
labs(x="opposing submission content",y="Authors involved in threads", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$unique_authors ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$unique_authors by thread_data$opposing
t = 2, df = 43, p-value = 0.05
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
0.0121 5.8046
sample estimates:
mean in group 0 mean in group 1
8.38 5.47 ANOVA for opposing vs. congruent x wave
summary(aov(unique_authors ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 132 132.4 2.13 0.15
wave 1 22 22.3 0.36 0.55
opposing:wave 1 15 15.4 0.25 0.62
Residuals 104 6462 62.1 ANOVA for opposing vs. congruent x ideology (subreddit)
summary(aov(unique_authors ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 132 132.4 2.16 0.14
subreddit 1 121 121.5 1.98 0.16
opposing:subreddit 1 4 4.3 0.07 0.79
Residuals 104 6374 61.3 3-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
summary(aov(unique_authors ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 132 132.4 2.18 0.143
subreddit 1 121 121.5 2.00 0.161
wave 1 0 0.3 0.01 0.943
opposing:subreddit 1 4 4.3 0.07 0.790
opposing:wave 1 24 24.2 0.40 0.530
subreddit:wave 1 244 244.5 4.02 0.048 *
opposing:subreddit:wave 1 21 21.4 0.35 0.555
Residuals 100 6084 60.8
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Deliberative Structure / Complexity –> Users
Creating measure for further analyses / comparison with deliberative quality: deliberative structure. unique_authors x max_depth
data <- data%>%
mutate(del_complexity = max_thread_depth*unique_authors)
thread_data <- data %>%
group_by(id_sub) %>% slice(1)
thread_data <- data.frame(thread_data)par(mar = c(4, 4, .1, .1))
# comment level data
my.mean<-tapply(data[,"del_complexity"],data[,"subreddit"],mean, na.rm =T)
my.sd<-tapply(data[,"del_complexity"],data[,"subreddit"],sd)
my.nr<-as.vector(table(data[,"subreddit"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("climate change","climate skeptics")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| climatechange | climate change | 122 | 92.3 | 3.58 |
| climateskeptics | climate skeptics | 191 | 213.0 | 7.83 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="deliberative complexity of threads",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
my.mean<-tapply(thread_data[,"del_complexity"],thread_data[,"subreddit"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"del_complexity"],thread_data[,"subreddit"],sd)
my.nr<-as.vector(table(thread_data[,"subreddit"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("climate change","climate skeptics")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| climatechange | climate change | 61.2 | 74.8 | 11.4 |
| climateskeptics | climate skeptics | 49.5 | 105.7 | 13.1 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue3")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="deliberative complexity of threads", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
df <- thread_data %>%
dplyr::group_by(subreddit, wave, opposing)%>%
dplyr::summarise(se = sd(del_complexity)/sqrt(length(del_complexity)),
mean = mean(del_complexity))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 21.08 | 95.4 |
| climatechange | australia | 1 | 15.12 | 31.2 |
| climatechange | california | 0 | 9.46 | 36.1 |
| climatechange | california | 1 | 11.71 | 20.0 |
| climateskeptics | australia | 0 | 9.94 | 26.1 |
| climateskeptics | australia | 1 | NA | 138.0 |
| climateskeptics | california | 0 | 19.18 | 61.7 |
| climateskeptics | california | 1 | 5.77 | 11.1 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("seagreen", "seagreen3"))+
labs(x="opposing submission content",y="deliberative complexity of threads", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$del_complexity ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$del_complexity by thread_data$opposing
t = 3, df = 82, p-value = 0.01
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
7.79 62.02
sample estimates:
mean in group 0 mean in group 1
60.3 25.4 ANOVA for opposing vs. congruent x wave
summary(aov(del_complexity ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 19076 19076 2.14 0.15
wave 1 6361 6361 0.71 0.40
opposing:wave 1 1677 1677 0.19 0.67
Residuals 104 925760 8902 ANOVA for opposing vs. congruent x ideology (subreddit)
summary(aov(del_complexity ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 19076 19076 2.14 0.15
subreddit 1 5905 5905 0.66 0.42
opposing:subreddit 1 1281 1281 0.14 0.71
Residuals 104 926613 8910 3-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
summary(aov(del_complexity ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 19076 19076 2.19 0.142
subreddit 1 5905 5905 0.68 0.412
wave 1 2978 2978 0.34 0.560
opposing:subreddit 1 1306 1306 0.15 0.700
opposing:wave 1 3545 3545 0.41 0.525
subreddit:wave 1 23149 23149 2.66 0.106
opposing:subreddit:wave 1 25318 25318 2.90 0.091 .
Residuals 100 871597 8716
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Gonzalez Width
ggplot(data, aes(max_thread_depth, gonzalez_width))+
geom_point(aes(colour = as.factor(opposing), size = n_comments_sub),
position = position_jitter(width = 2, height = 5),
alpha = 0.5)+
geom_point(data = subset(data, opposing == 1),
aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 2, height = 5),
alpha = 0.5)+
#theme_grey()+
facet_grid(vars(wave), vars(subreddit))+
xlab("Depth")+
ylab("Width (Gonzalez-Bailon, 2010)")+
scale_colour_manual(name="Type",
labels=c("congruent", "opposing"),
values=c("#2E8B57", "indianred")) +
scale_size_continuous(name="N comments")
par(mar = c(4, 4, .1, .1))
# comment level data
my.mean<-tapply(data[,"gonzalez_width"],data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(data[,"gonzalez_width"],data[,"opposing"],sd)
my.nr<-as.vector(table(data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 8.01 | 5.45 | 0.154 |
| 1 | opposing | 3.37 | 2.09 | 0.171 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="Maximum width of comment thread at any given depth",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
thread_data <- data %>%
group_by(id_sub) %>% slice(1)
thread_data <- data.frame(thread_data)
my.mean<-tapply(thread_data[,"gonzalez_width"],thread_data[,"opposing"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"gonzalez_width"],thread_data[,"opposing"],sd)
my.nr<-as.vector(table(thread_data[,"opposing"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("congruent","opposing")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| 0 | congruent | 3.48 | 4.11 | 0.436 |
| 1 | opposing | 1.79 | 2.10 | 0.481 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="Maximum width of comment thread at any given depth", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
df <- thread_data %>%
group_by(subreddit, wave, opposing)%>%
summarise(se = sd(gonzalez_width)/sqrt(length(gonzalez_width)),
mean = mean(gonzalez_width))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 0.801 | 5.25 |
| climatechange | australia | 1 | 1.393 | 3.20 |
| climatechange | california | 0 | 0.514 | 2.54 |
| climatechange | california | 1 | 0.632 | 1.00 |
| climateskeptics | australia | 0 | 0.512 | 1.92 |
| climateskeptics | australia | 1 | NA | 4.00 |
| climateskeptics | california | 0 | 0.764 | 3.42 |
| climateskeptics | california | 1 | 0.441 | 1.12 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("steelblue4", "steelblue2"))+
labs(x="opposing submission content",y="G Width of threads", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$gonzalez_width ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$gonzalez_width by thread_data$opposing
t = 3, df = 52, p-value = 0.01
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
0.392 2.996
sample estimates:
mean in group 0 mean in group 1
3.48 1.79 ANOVA for opposing vs. congruent x wave
summary(aov(gonzalez_width ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 45 44.9 3.05 0.084 .
wave 1 24 23.8 1.61 0.207
opposing:wave 1 8 8.0 0.55 0.462
Residuals 104 1534 14.7
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1ANOVA for opposing vs. congruent x ideology (subreddit)
summary(aov(gonzalez_width ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 45 44.9 3.04 0.084 .
subreddit 1 27 26.8 1.82 0.181
opposing:subreddit 1 1 0.8 0.05 0.817
Residuals 104 1538 14.8
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 13-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
summary(aov(gonzalez_width ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 45 44.9 3.12 0.080 .
subreddit 1 27 26.8 1.87 0.175
wave 1 10 9.8 0.68 0.410
opposing:subreddit 1 1 0.8 0.06 0.810
opposing:wave 1 12 11.6 0.81 0.372
subreddit:wave 1 65 64.6 4.49 0.036 *
opposing:subreddit:wave 1 14 13.6 0.95 0.333
Residuals 100 1438 14.4
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Deliberative Structure / Complexity –> Gonzalez Width
Creating measure for further analyses / comparison with deliberative quality: deliberative structure. gonzalez_width x max_depth
data <- data%>%
mutate(del_complexity = max_thread_depth*gonzalez_width)
thread_data <- data %>%
group_by(id_sub) %>% slice(1)
thread_data <- data.frame(thread_data)
ggplot(data, aes(max_thread_depth, gonzalez_width))+
geom_point(aes(colour = as.factor(opposing), size = n_comments_sub),
position = position_jitter(width = 3, height = 3),
alpha = 0.5)+
geom_point(data = subset(data, opposing == 1), aes(colour = as.factor(opposing), size = ups_sub),
position = position_jitter(width = 6, height = 0.08),
alpha = 0.5)+
theme_minimal()+
facet_grid(vars(wave), vars(subreddit))+
xlab("Depth")+
ylab("Width")+
scale_colour_manual(name="Type",
labels=c("congruent", "opposing"),
values=c("#2E8B57", "indianred")) +
scale_size_continuous(name="Number of Comments")
ggsave(file="../../output/G_width_depth.png",width = 8, height = 6, dpi = 500)par(mar = c(4, 4, .1, .1))
# comment level data
my.mean<-tapply(data[,"del_complexity"],data[,"subreddit"],mean, na.rm =T)
my.sd<-tapply(data[,"del_complexity"],data[,"subreddit"],sd)
my.nr<-as.vector(table(data[,"subreddit"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("climate change","climate skeptics")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| climatechange | climate change | 55.1 | 40.5 | 1.57 |
| climateskeptics | climate skeptics | 91.7 | 94.7 | 3.48 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="deliberative complexity of threads",title="Comment level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data
my.mean<-tapply(thread_data[,"del_complexity"],thread_data[,"subreddit"],mean, na.rm =T)
my.sd<-tapply(thread_data[,"del_complexity"],thread_data[,"subreddit"],sd)
my.nr<-as.vector(table(thread_data[,"subreddit"]))
my.se<-my.sd/sqrt(my.nr)
my.df<-data.frame(groups=factor(names(my.mean),labels =c("climate change","climate skeptics")),
mean=my.mean,
sd=my.sd,
se=my.se)
my.df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| groups | mean | sd | se | |
|---|---|---|---|---|
| climatechange | climate change | 27.4 | 34.7 | 5.29 |
| climateskeptics | climate skeptics | 23.0 | 51.1 | 6.33 |
ggplot(my.df, aes(x=groups, y=mean)) +
geom_bar(stat="identity", aes(x=groups, y=mean),size=0.1,
width = 0.6, fill=c("steelblue3")) +
theme(
axis.text.x = element_text(colour="black", size =12),
axis.text.y = element_text(colour="black", size =10),
axis.title.x = element_text( colour="black", size=12),
axis.title.y = element_text( colour="black", size=12))+
labs(x="",y="deliberative complexity of threads", title="Thread level data")+
geom_errorbar(aes(ymin = mean-my.se, ymax = mean+my.se), width=.1)
# thread level data + grouped for subreddits
df <- thread_data %>%
dplyr::group_by(subreddit, wave, opposing)%>%
dplyr::summarise(se = sd(del_complexity)/sqrt(length(del_complexity)),
mean = mean(del_complexity))
df%>%
xtable()%>%
kable()%>%
kable_styling(bootstrap_options = "striped",
full_width = FALSE,
position = "center")| subreddit | wave | opposing | se | mean |
|---|---|---|---|---|
| climatechange | australia | 0 | 9.61 | 45.30 |
| climatechange | australia | 1 | 7.76 | 15.60 |
| climatechange | california | 0 | 3.31 | 12.62 |
| climatechange | california | 1 | 4.67 | 6.40 |
| climateskeptics | australia | 0 | 5.83 | 13.23 |
| climateskeptics | australia | 1 | NA | 92.00 |
| climateskeptics | california | 0 | 9.15 | 28.02 |
| climateskeptics | california | 1 | 1.35 | 2.88 |
ggplot(df, aes(x=opposing, y=mean, fill = subreddit)) +
geom_bar(position=position_dodge(), stat="identity", colour='white') +
geom_errorbar(aes(ymin=mean-se, ymax=mean+se), width=.1,position=position_dodge(.9))+
theme_minimal()+
scale_fill_manual(values=c("seagreen", "seagreen3"))+
labs(x="opposing submission content",y="deliberative complexity of threads", title="Thread level data")+
facet_grid(vars(wave))
T-Test for opposing vs. congruent submissions
# overall t-test
t.test(thread_data$del_complexity ~ thread_data$opposing)
Welch Two Sample t-test
data: thread_data$del_complexity by thread_data$opposing
t = 2, df = 59, p-value = 0.03
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
1.14 30.16
sample estimates:
mean in group 0 mean in group 1
27.5 11.8 ANOVA for opposing vs. congruent x wave
summary(aov(del_complexity ~ opposing + wave + opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 3836 3836 1.90 0.17
wave 1 2923 2923 1.45 0.23
opposing:wave 1 865 865 0.43 0.51
Residuals 104 210373 2023 ANOVA for opposing vs. congruent x ideology (subreddit)
summary(aov(del_complexity ~ opposing + subreddit + opposing * subreddit, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 3836 3836 1.87 0.17
subreddit 1 933 933 0.46 0.50
opposing:subreddit 1 356 356 0.17 0.68
Residuals 104 212871 2047 3-way ANOVA for opposing vs. congruent x ideology (subreddit) x wave
summary(aov(del_complexity ~ opposing + subreddit + wave + opposing * subreddit + opposing * wave + subreddit * wave +
subreddit * opposing * wave, data = thread_data)) Df Sum Sq Mean Sq F value Pr(>F)
opposing 1 3836 3836 1.97 0.164
subreddit 1 933 933 0.48 0.491
wave 1 2105 2105 1.08 0.301
opposing:subreddit 1 368 368 0.19 0.665
opposing:wave 1 1643 1643 0.84 0.361
subreddit:wave 1 4850 4850 2.49 0.118
opposing:subreddit:wave 1 9264 9264 4.75 0.032 *
Residuals 100 194998 1950
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1